fooker

7 ms·

121.
by fooker 2mo ago
Yeah it's for simulating how many neutrons are going to be in the air. And what fraction of electromagnetic waves and particles are going to absorbed by those neutrons and other things in the air. Pretty much weather!
122.
by fooker 2mo ago
Mathematicians routinely spend years on a problem without getting anywhere.
123.
by fooker 2mo ago
This misses a very important point. It’s not just about out remembering, it’s about breadth. Mathematicians are all about depth. It’s pretty much impossible to become an expert in more than one narrow field of mathematics. AI is happily app
124.
by fooker 2mo ago
k
125.
by fooker 2mo ago
> you think kids in other states Parents of the kids, realistically.
126.
by fooker 2mo ago
We are not getting that reckoning. It’s great to yearn for bug free software, but that necessarily brings an insane amount of red tape to get anything done. Like 5 years worth of review to start building your code with the next version of y
127.
by fooker 2mo ago
> reword my initial prompt to get the agent off an unintended track. The signal here is the action of stopping the agent “do something else, this is stupid”, not a tweak to the initial prompt.
128.
by fooker 2mo ago
Okay yeah, fair point. My comment was from a decade old perspective
129.
by fooker 2mo ago
Google has close to the best internal tooling in the industry for a decade or so. Then the Google engineers who joined Facebook missed it so much that they built a better replacement.
130.
by fooker 2mo ago
There's a full fledged 'reasoning' step that basically expands your prompt. As long as you are not missing important information, how you word the prompt does not have any effect.
131.
by fooker 2mo ago
There's no program you run to 'make the thing'. It's mostly ad hoc scripts and some pretty horrible hacks being run by a hundred engineers trying to improve a thousand different things at once with a hundred thousand GPU
132.
by fooker 2mo ago
This trope was valid maybe in 2022. Model training now is not a straight forwards process of input data -> run tools -> get model. There's a whole lot of alchemy going on. We don't quite understand what works and what doesn&
133.
by fooker 2mo ago
Exact prompts haven't mattered for about a year now.
134.
by fooker 3mo ago
Exactly right. Not only is it specialized, it also has its own field specific memes* and trends that evolve over time. The purpose is mostly to signal "I'm one of you". *memes as in the actual meaning of memes, not internet m
135.
by fooker 3mo ago
I think, maybe you misunderstood my comment? I'm all for LLMs producing good research. That's obviously happening right now. What I said was people with good research being penalized for not knowing in-group language. It's no
136.
by fooker 3mo ago
Publishing CS papers at the top venues requires using in-group language and formalism that is pretty much inaccessible to someone who has not done a PhD in that specific narrow field. LLMs are pretty good at this, hilariously. This means a
137.
by fooker 3mo ago
How's that asymmettry worse than it is now?
138.
by fooker 3mo ago
Why would you assume that the same capability could not be used to harden systems?
139.
by fooker 3mo ago
Anthropic's fall from grace and mindshare seems rather accelerated. I wonder if these rapid movements are going to be the norm now. I imagine there would be angry investors if this sort of thing happened with a public company.
140.
by fooker 3mo ago
Sure, agreed. The question then is, how much memorization is really needed for intelligence if you can query structured information?
141.
by fooker 3mo ago
I'm claiming that the max amount of intelligence you get out of a 100GB model is unlikely to be that much lower than what you can get out of a 1TB model.
142.
by fooker 3mo ago
No, it's like saying a chip that consumes 500W is going to be better than a chip that consumes 200W. This used to be true for the first several decades of microprocessors, and stopped being true in the early 2000s.
143.
by fooker 3mo ago
> Some groups are baking models into silicone While some other groups are baking silicone into models :)
144.
by fooker 3mo ago
Prediction - we are going to figure out SOTA AI performance without requiring 1TB of memory within a year or so. Of course CXMT, Micron, and family will still be profitable, but maybe not 'surge 470% from IPO' profitable.
145.
by fooker 3mo ago
Great, so the other member of the set matters for you more than cost. Do you actually need to run the state of art model at 5 tokens per second instead of a qwen or whatever 7b or 30b model at 100 tokens per second?
146.
by fooker 3mo ago
> Even if the output is like 5-6 tok/s, that might be usable for some purposes. You'll spend ~100x more on electricity than the API cost to have it run on someone else's GPU at several hundred tokens per second. I think s
147.
by fooker 3mo ago
agentic flimble please.
148.
by fooker 3mo ago
It's a basic step to make a plumbis
149.
by fooker 3mo ago
Not that you have much reason to believe a stranger on the internet, but I have seen this in person. Not doing acrobatics like the marketing video of course, just doing mundane things like going from point A to point B.
150.
by fooker 3mo ago
Reminds me of the NYC cartoon too ;)

More ›